In mountainous terrain, where the wind power potential is largest, the estimation of the local wind power can be done rationally by means of available information about the large-scale flow and the detailed terrain and numerical flow models for downscaling, provided that the numerical model estimates can be assigned sufficient confidence. In this study the confidence of a local model in such an estimation system is discussed. The model is based upon the Reynolds-averaged Navier–Stokes equations with (K, ϵ) turbulence closure and integrated with finite element numerical techniques. The model has previously been validated relative to complicated laboratory-scale flows and appears to predict some full-scale geophysical flows plausibly. Here its predictions are compared quantitatively with the full-scale Askervein hill experimental data. The model estimates the data to within the experimental uncertainty, which we judge to be comparable to 10%, as other comparable models also do. This contributes to assign confidence to the downscaling estimation system mentioned. Copyright © 2004 John Wiley & Sons, Ltd.